The Audience Is Dead. Long Live the Decision.
This Signal & Noise exclusive is brought to you by AUDIENCES.
For decades, advertising has been built around audiences: define the segment, move it into a platform, and optimize against it. AI is starting to invert that model, shifting the focus from who someone is to what a brand should do next based on fresh customer context. In that world, the audience doesn’t disappear—but it becomes the output of a decision rather than the center of the strategy.
For most of modern advertising, an audience has been a container we stuff people into. We take people who share some characteristic, behavior, or inferred intent and put them into a group—recent purchasers, high-value customers, cart abandoners, auto intenders, frequent travelers, new parents, people likely to churn. Then, of course, we decide what to do with the group.
That model made sense when marketing technology was built around lists, segments, batch processing, and relatively blunt activation systems. Given this ecosystem, it was the job of the marketer to find the right people, put them into the right audience, and send that audience somewhere it could be activated.
But AI is beginning to make this model look primitive. The most important question is no longer "Which audience does this customer belong to?" Increasingly, it is "What should I do with this customer right now?"
Should I acquire them?
Retain them? Suppress them?
Upsell them?
Show them a different product?
Increase the bid?
Decrease it?
Change the message?
Do nothing at all?
I would argue these are decisions, not segments. And this distinction matters. Based on preexisting data, a customer can belong to the same audience on Monday and Friday, while the correct marketing decision changes completely based on facts on the ground. On Monday, for example, someone browsing a new laptop might be a high-value prospect worth aggressively pursuing. By Wednesday, however, they may have purchased the laptop directly from the brand. On Friday, they may be researching accessories, returning the product, contacting customer service, or shopping for something entirely different.
In a CRM, CDP, or media platform, the label might still say “high-value technology shopper.” But that label tells us very little about what the marketer should actually do next.
This is where the traditional audience model begins to break down. In real life, customers are not static. Their relationship with a company changes constantly. They browse, purchase, abandon, return, subscribe, cancel, upgrade, complain, engage, disengage, and come back again. Yet much of advertising still takes this continuously changing customer and stuffs them into a relatively static bucket.
My proposition is that AI gives us an opportunity to invert this model. Instead of building an audience first and deciding what to do with it second, we can increasingly start with the decision. This means given everything we know about this customer—their behavior, purchases, preferences, consent, value, context, and recent interactions—we need to decide the best action to take now. The audience then becomes the output of that decision.
You may think this sounds like a semantic distinction, but I think it represents a much larger change in how marketing systems will work. For decades, the audience has been the central unit of activation.
In the next generation of marketing, the central unit may be the decision itself.
The Segment Isn't Wrong—it's Just Incomplete
To be clear, I don't think segmentation is going away anytime soon. There will still be tremendous value in understanding whether someone is a high-value customer, a frequent traveler, a recent purchaser, a likely churner, or an in-market auto buyer. These classifications tell us something useful about the customer.
The problem comes when we mistake the classification for the action. Knowing that I am a “frequent traveler,” for example, does not tell an airline what it should do with me today. I might be shopping for a flight, or I might have just booked one. I might be sitting in an airport after my flight was canceled, or I might have recently achieved a new loyalty tier. I might not be planning to travel again for six months. At the end of the day, I am still a frequent traveler in every one of those scenarios. But the correct decision is completely different.
This is why I think audiences increasingly become inputs into a decisioning process, rather than the final output of the marketing process itself. This means a decision might start with the fact that someone is a high-value customer, but it should also incorporate what that person has done recently—what they have purchased, what they are currently browsing, what communications they have already received, whether they have an unresolved service issue, what permissions they have granted, how valuable they are likely to be in the future, and potentially hundreds of other signals.
Critically, many of these signals have very different shelf lives. The fact I've been a customer for 10 years might remain relevant for years, but the fact that I bought a laptop may change the appropriate marketing action immediately. A search I performed twenty minutes ago may matter enormously right now... and almost not at all three months from now. You get my drift.
Traditional segmentation tends to flatten all of this context into a label. Good decisioning can preserve the context. Think again about our laptop shopper. A traditional audience workflow might identify this person as “in-market for laptops,” push that segment into several media platforms, and begin targeting them aggressively with laptop ads that follow them around the Internet.
The problem is the media machine often keeps running even after reality changes:
The customer buys the laptop.
The audience persists.
The ads keep coming, and coming, and coming.
Anyone who has bought something online and then spent the next two weeks being followed around the internet by ads for the exact product they just purchased has experienced this annoying problem firsthand. We tend to describe this as an audience synchronization problem, a suppression problem, or perhaps a media waste problem.
But I think there's a deeper underlying issue: The system is optimizing against a classification instead of responding to a decision. Once the customer buys the laptop, the relevant question should immediately change. We no longer need to ask whether this person belongs in the “laptop shopper” audience.
Instead, we need to ask what should happen next.
Maybe they should be suppressed from ALL laptop acquisition campaigns. Maybe they're now a strong candidate for a monitor, docking station, warranty, or software subscription. Maybe the right action is simply to leave them alone for a while. And if they opened a customer-service case the next morning, perhaps the most intelligent marketing decision is to stop selling them anything until the issue is resolved.
None of those decisions require us to abandon audiences. They require us to stop treating the audience as the end of the process. In other words, the audience tells us who someone is likely to be. The decision tells us what we should do about it now.
Decisioning Requires Fresh Signals
There is, however, a pretty obvious catch to all of this. You cannot make a good decision using bad—or stale—information. If the goal is to determine the best thing to do with a customer right now, then the underlying signals informing that decision also need to reflect what is really happening right now. This sounds obvious, but it runs directly into the way a lot of marketing infrastructure still works.
Much of the traditional advertising ecosystem was not designed for immediacy. Data gets collected in one system, processed in another, transformed somewhere else, loaded into a customer platform, pushed into an audience, synced to a media platform, matched against an identity graph, and eventually made available for activation.
By the time the signal arrives, the customer may have moved on. This creates a new way to think about the value of customer data: not every signal decays at the same rate. Some customer facts have a very long shelf life, for example my home market, loyalty status, historical lifetime value, or broad product affinity might remain useful for months or years. Other signals, by contrast, decay incredibly quickly. If I searched for a hotel in London 10 minutes ago, that might be highly valuable information. If I searched for one nine months ago, probably not so much.
If I just bought a new television, that fact should immediately influence whether you continue bidding to sell me another television. If I just filed an angry customer-service complaint, there is a decent chance the last thing I want to see is an upbeat ad telling me how much your company values my business. This is why, in a decision-driven marketing model, latency becomes a strategic issue. It isn't enough to have the right data stored somewhere inside the enterprise. The relevant signal has to become available to the decisioning and activation systems quickly enough to change what they do.
This changes how we should think about first-party data. For years, marketers have talked about building the most complete possible view of the customer—a golden record, a Customer 360, a single source of truth, and so on. Of course these things are still valuable, but completeness alone isn't enough. A perfectly complete view of the customer that is five days old may be considerably less useful for advertising than a smaller set of signals that tells me what changed five minutes ago.
Increasingly, the advantage will come from having the right context at the right moment, not simply having the largest possible customer profile. And this is where AI raises the stakes even further, because AI can process vastly more information and make vastly more decisions quicker than a human marketer ever could. But it can also make the wrong decision at extraordinary speed if the information feeding it is stale.
Giving an AI system last week's customer context doesn't magically produce intelligent marketing.
If anything, it helps you automate yesterday's decision faster.
This is why I think the winners in this next era of marketing won't just be the companies with the most customer data, but those capable of turning changes in customer behavior into usable signals quickly enough for those signals to affect the next decision.
Once marketing becomes a decisioning problem, data freshness becomes part of the decision itself.
The Decision Should Travel, Not the Customer Data
This creates another problem, however. If marketers need fresher customer context to make better decisions, the obvious answer might seem to be: move more customer data, more frequently, into more places. I think this is the wrong conclusion.
The advertising ecosystem already has a data movement problem. Customer information gets copied from warehouses into CDPs, from CDPs into onboarding platforms, from onboarding platforms into media platforms, from media platforms into measurement environments, and often into plenty of other systems along the way. Every additional copy introduces latency, complexity, cost, governance challenges, and another place where potentially sensitive customer information needs to be controlled.
If we're moving toward a world of continuous decisioning, we shouldn't respond by building an even bigger network of customer-data plumbing. Instead, we should increasingly move the decision—or the signal required to execute the decision—rather than the underlying customer data itself. This is a very important distinction.
Imagine a retailer determines that a particular customer has just purchased a laptop. A media platform doesn't necessarily need to know the customer's complete purchase history, lifetime value, loyalty profile, browsing history, customer-service interactions, and everything else the retailer knows about that person.
It may simply need to know: Stop bidding on this person for laptop acquisition.
Or perhaps: This customer is now eligible for an accessories campaign.
The intelligence that produced the decision can remain close to the company's first-party data. What travels downstream is the result. This is a fundamentally different architecture from the one much of advertising was built around.
Historically, we've tended to push as much customer information as possible toward the activation platform and let the platform figure out what to do with it. In a decision-driven model, more of the intelligence happens upstream, where the company has the richest context about the customer. The media platform then becomes less responsible for defining the customer and more responsible for executing the decision.
This has some significant advantages. First, it reduces how much customer data needs to move in the first place. It also allows brands to maintain greater control over their proprietary customer intelligence, and can reduce latency between a change in customer behavior and the resulting marketing action. It also prevents every advertising platform from developing a slightly different interpretation of who the customer is and what they're worth.
Most importantly, it keeps the brand—not the media platform—at the center of the decision. Meta may know a tremendous amount about how someone behaves on Meta, and Amazon may understand what they do on Amazon. Google has its own enormous universe of signals. These are incredibly valuable inputs, but none of these Walled Gardens has the complete relationship that a company has with its own customer.
At the end of the day, the brand knows that I bought the product yesterday. It also knows I'm a platinum loyalty member, and I returned my last two orders. It knows I opened a service ticket this morning, and it should know what I'm worth to the business. Increasingly, it can use AI to understand how those facts should influence what happens next.
This intelligence shouldn't have to be reconstructed independently inside every destination where the brand buys media. The brand should make the decision once, and the decision should travel.
The Media Platform Doesn't Disappear—Its Job Changes
No, none of this means the media platforms become less important. In fact, I would argue the opposite.
Google, Meta, Amazon, TikTok, DSPs, retail media networks, and other activation environments possess enormous amounts of intelligence that brands simply do not have. They understand what is happening inside their own ecosystems at a level no advertiser possibly could.
They know which inventory is available, and they understand contextual signals. They see engagement patterns across millions or billions of interactions, and they can predict the likelihood that someone will click, watch, convert, or ignore an ad. Increasingly, their AI systems can make these calculations in fractions of a second across an almost incomprehensible number of opportunities. Brands need this intelligence.
What changes is the division of labor. Historically, advertisers have often handed a media platform an audience and effectively said: Here are the people I think I care about. Now go figure out which of them to reach, what they're worth, and how much I should pay. The platform then uses its own signals and optimization algorithms to make those decisions.
In the model I'm describing, the brand arrives with considerably more intelligence of its own. It might know that one customer is worth significantly more than another, or that another customer just purchased the product being advertised. A third might be at serious risk of churn, and a fourth may have just become eligible for a completely different offer.
The platform still determines whether an impression is attractive, what inventory is available, what the contextual environment looks like, and potentially what bid or creative is most likely to perform. But now it can make those decisions with a much better signal from the advertiser about what the customer actually means to the business.
This distinction is important because the two sides know different things. A media platform may be exceptionally good at answering "What is happening here?" This includes what content is this person consuming, what device are they using, how likely are they to engage, and what the impression costs.
The brand, meanwhile, is better positioned to answer "What does this customer mean to us?" Did they purchase yesterday? Are they profitable? Are they loyal? Are they at risk of leaving? Do they already own the product, or did they just complain to customer service? Are they eligible for this offer in the first place? The reality is neither side has the complete answer on its own. The opportunity is thus to combine them at the moment the decision needs to be made.
This is one reason I think the next generation of advertising will be less about building bigger audience files and more about creating a faster exchange of intelligence between the enterprise and the activation environment. This implies instead of saying, “Here is my audience,” the brand increasingly says something more like:
Here is what I know.
Here is what I want to accomplish.
Here is the decision we've made about this customer.
Now use what you know to execute it as effectively as possible.
Thus a very different relationship, and one that changes where competitive advantage lives. If every advertiser ultimately has access to many of the same media platforms, increasingly similar AI capabilities, and much of the same inventory, then the differentiator cannot simply be the technology used to buy the impression.
The advantage comes from the proprietary intelligence you bring to that impression:
Your customer relationships.
Your transaction history.
Your product usage.
Your loyalty data.
Your service interactions.
Your understanding of customer value.
And, increasingly, your ability to turn all of these signals into a useful decision before the media platform ever has to decide whether to bid. This is why I don't think the solution is about replacing media-platform intelligence with brand intelligence. Instead, it's about combining the two. The media platform remains extraordinarily good at understanding the opportunity in front of it. The brand becomes much better at telling the platform what that opportunity is actually worth.
The Audience Infrastructure Has to Change
If I'm correct and this is where advertising is headed, then the infrastructure connecting customer data to media certainly has to change along with it. The fact is, most audience technology we rely on today was built for a world in which the primary objective was to create a segment and move that segment somewhere else. Identify the people, package them into an audience, match them to an identifier, push the audience into a platform, and activate it.
Sure, this workflow made perfect sense when audiences themselves were the actual product. But it makes considerably less sense when the real objective is continuous decisioning. If a customer's behavior changes at 10:02am, I don't necessarily want to wait for another batch process, another audience refresh, another identity match, and another platform sync before that change can influence what happens at 10:03am.
Ultimately, the infrastructure needs to become more responsive to the underlying customer intelligence.
This means bringing activation closer to where the first-party data already lives, reducing unnecessary movement of customer information, and making it easier to translate changes in customer context into signals that media platforms can actually use.
FWIW, this is an area where our partners at AUDIENCES have been thinking about the problem differently. Their approach starts from a fairly simple premise: brands shouldn't have to continuously extract and replicate massive amounts of customer data just to make that intelligence useful in advertising. Instead, more of the audience logic happens closer to the company's own data environment, with the relevant output then made available for activation across media destinations.
I think this is important because it shifts the purpose of audience technology. The goal is no longer simply to answer: "Who belongs in this segment?" Increasingly, it needs to help answer:
What has changed, what does that change mean, and where does that decision need to go?
That may still result in an audience—in many cases, it probably will. But the audience becomes a delivery mechanism for intelligence rather than the intelligence itself.
This distinction becomes increasingly important as AI enters the picture. An AI-driven marketing system can constantly evaluate customer behavior, transaction history, loyalty status, product ownership, propensity, consent, service interactions, and dozens—or hundreds—of other signals. It can determine that a customer's value has changed, their intent has shifted, or that an entirely different action is now appropriate.
In this paradigm, the interesting question is what happens next. Can that change influence activation quickly? Can it do so without creating another unnecessary copy of the underlying customer data? Can the brand preserve control over the logic that produced the decision? And, most importantly, can that intelligence be applied consistently across multiple media environments rather than reconstructed independently inside every platform?
In the future, these questions are going to matter a lot more as marketing moves from periodic audience creation toward continuous customer decisioning. Because ultimately, the objective isn't to build a better bucket, but rather to construct a system that knows when the bucket no longer matters.
The Audience Isn't Dead—It's Been Demoted
So, despite the title of this article, the audience isn't actually dead. I was being hyperbolic. We're going to keep building audiences, keep activating them, and media platforms will continue to ingest them, optimize against them, and measure their performance. There is simply too much existing infrastructure—and too much genuine utility—for the audience to disappear anytime soon.
But I do think the audience's role is changing. For much of the digital advertising era, the audience has sat squarely at the center of the process. We collected data in order to build an audience, moved that audience into a platform, bought media against it, measured the results, and went back and built another audience. The audience was both the abstraction and the main event.
As we look to the future, AI starts to separate those two things. In many ways, the abstraction may still be useful. I still want to know who my high-value customers are. I still want to understand which customers are likely to churn, which prospects are in market, and which buyers have an affinity for a particular product.
But increasingly, the action should be determined by something more dynamic.
What do I know about this person right now?
What has changed?
What is the business trying to accomplish?
What are the constraints?
Given all of this information, what is the smartest thing to do next? This is ultimately a decision.
The companies that become best at making these decisions—and getting them into market quickly—will have a meaningful advantage over the companies still relying primarily on static segments, periodic audience refreshes, and platform-level optimization to figure everything out for them.
This also changes the role of first-party data. The competitive advantage isn't simply possessing more of it. We've already spent years filling warehouses, lakes, CDPs, and clouds with customer information.
The advantage comes from being able to turn that information into a better decision while the information still matters. And increasingly, AI gives us the ability to do that at a scale and speed that simply wasn't practical before.
So perhaps the audience isn't dead after all. Maybe it has simply been demoted in a future in which it is no longer necessarily the place where the intelligence lives. It becomes one of the mechanisms through which intelligence gets executed.
The customer changes, signals change, the decision changes. And the audience follows. Long live the decision.
------------------------------------------------
Rio is an executive with 20+ years at the intersection of strategy consulting, AdTech, data, and media. He's a trusted advisor on customer experience, digital strategy, and marketing transformation. He's a partner at Credera, Omnicom's consulting arm. He's also a podcast host, writer, and public speaker focused on the future of advertising and AI-driven infrastructure.
Comments